OCR - Optical character recognition
AI - Artificial intelligence
I mean, this method - using other people answers is really good then you need to recognize printed text.
The main problem of OCR is to "teach" program engine to recognize printed text, because there is tons of different fonts, and not all characters printed or scanned fine. In basic, there is some workers, who manually recognize characters when program fails, and tells the program which those character was. Program "learns" and recognizing text even better. This people not work for "thanks", though.
The method I told about used by ReCAPTCHA
, CAPTCHA providing internet service. Their CAPTCHA showing not one word, but two - one is the control, other one is the word current engine can't recognize. User can't say which word is control, so, he needs to recognize and enter both of them (actually, on this moment of time and after great engine improvement skilled user can define which word is control), which is serious economy on full-day workers. Em, actually, they just automatically paste recognized word into the text (and will make after-check, I hope).- Test it yourself -
The problem of using this method in Cleverbot
- program analysing formalities, not meaning of statements (which is really hard, I told you). Good enough for IRC chat, bad in Turing tests, cause in IRC you wouldn't suspect others they are AI (all the time, though).